Cyrille Rossant

Regarding fullscreen: this is not implemented yet (note: I’ve never needed it so far, as I mostly use i3 on Linux, which enables the fullscreen toggle on all windows). It...

Regarding the texture update warnings: I'll have a look. Do you see them even without activating the Vulkan SDK validation warnings/errors?

It seems you can work around the errors at least (but not the warnings) by adding in `renderer.cpp` on `L775`: ```c dvz_gpu_wait(rd->gpu); ``` This is a really ugly hack but...

Ok, for the fullscreen feature you'll mostly want to see `glfw_utils.h` and `window.h`

I think the figure-related code is in `scene.c` (which should be refactored at some point...). Regarding the `app.xxx(figure...)` pattern elsewhere in the code, although not strictly necessary, it anticipates a...

Oh, there will always be a single app instance. It's basically a singleton. We should enforce this in the code. The only reason there's an app is to avoid module...

There is still a distinction between the scene, the app, and the figure in the C API. The new Python API hides this from the user, but the underlying abstractions...

> I also recognize things are developing as we go, so not everything is decided or planned ahead in detail. (That's probably very common in these kinds of projects.) Yes,...

Instead of a timer, I'd rather use a window creation flag in the C API (`dvz_figure()`) that is passed to the renderer and that indicates the window should be created...

So the `request.c` file defines functions to prepare rendering requests sent to the renderer (`renderer.cpp`). The idea is that this protocol handles GPU rendering only, not windowing logic or user...